Statistics meets machine learning: the example of statistical boosting

Riccardo De Bin
Department of Mathematics
University of Oslo, Norway

Starting with a brief discussion on the relationship between statistics and machine learning (and data science) and on the basic concepts of statistical learning (overfitting, bias-variance trade-off, etc.), the workshop will focus on boosting, a specific method for data analysis which combines the powerfulness of a machine learning technique and the interpretability of a statistical model. Originally developed as a black-box for classification, boosting has been succesively statistically interpreted and, thanks to its flexibility, profitably applied to many statistical problems. Advantages of boosting, in addition to the aforementioned flexibility, include the ability to cope with high-dimensional data, implicit variable selection and shrinkage. The workshop will present boosting from a statistical point of view, focusing on its practical implementation in R and on the interepretation of the results.

The participants should be familiar with R programming and bring their own laptops, with the latest version of R. The use of R Studio is also encouraged. A wireless connection to the Internet will be provided.